Corporate Sport Sponsorship and Stock Returns

Corporate Sport Sponsorship and Stock Returns:
Evidence from the NFL
Assaf Eisdorfer
Elizabeth Kohl
April 2015
ABSTRACT
Most home stadiums/arenas of major-sport teams in the U.S. are sponsored by large publicly traded
companies. Using NFL data we find that stock returns to the sponsoring firms are affected by the
outcomes of games played in their stadiums. Wins in Monday night games generate next-day
abnormal returns 50 basis points higher than losses. The effect is 80 basis points in the post-season
and when the game outcome is unexpected. This does not revert over the next few days. Outcomes
of NFL games could serve as a reasonably exogenous instrument for investor sentiment.
Keywords: Stock returns; Sport sponsorship
JEL Classifications: A12, G12, G14
*
Assaf Eisdorfer is from University of Connecticut; [email protected].
Elizabeth Kohl is from University of Connecticut; [email protected]. We thank
Ivo Welch, and seminar participants at University of Connecticut, Hebrew University of
Jerusalem, and Interdisciplinary Center Herzliya for helpful comments. All errors are our own.
1. Introduction
Consider a major sport team that is hosting an important and highly rated televised match, at which
it is expected to win. The match is played at the team’s home stadium, which is named after the
team’s sponsor, a large and well-known publically traded corporation. The name of the sponsor is
therefore repeatedly mentioned and seen during the match. Now suppose that the team is
unexpectedly losing the match. Will this loss be reflected in the return on the sponsor’s stock in
the next trading day? If so, is it because the sponsor has been associated with the prevailing
disappointment in the team? Or is the market reacting to the financial implications of the loss – a
shorter season, fewer opportunities to promote the sponsor? Would the market have had a similar
reverse reaction after a win? In this study, we document that stock prices of companies that sponsor
National Football League (NFL) stadiums are affected by the outcomes of important individual
games played in the stadiums. Our evidence suggests that this effect is partially driven by investor
sentiment.
Several recent studies have explored the association between the outcomes of sport matches
and stock market returns. Boyle and Walter (2002) find no relation between the success of the New
Zealand national rugby team and the stock market reaction in the country. Ashton, Gerrard, and
Hudson (2003) find a strong relationship between the performance of the English national soccer
team and the change in the price of shares traded on the London stock exchange, where good (bad)
performances by the national team are followed by positive (negative) market returns. Edmans,
Garc'ia, and Norli (2007) conduct a cross-country analysis and find that losses in soccer (and other
sports) matches have an economically and statistically significant negative effect on the losing
country’s stock market; yet they find no evidence of a corresponding effect after wins. Scholtens
2
and Peenstra (2009) analyze matches of eight publically traded European soccer teams. They find
that the stock market response is significant and positive for victories and negative for defeats.
Our study is the first to examine the effect of professional sport match outcomes on stock
returns of the teams' sponsors.1 Although corporate sponsorship of professional sports stadia can
be traced back to the early 1900s, the number of major league teams playing in corporate-named
stadiums and arenas has sharply increased in the last two decades. As of the end of 2013, 62 percent
of the home stadiums/arenas of the four major league sports (football, baseball, basketball, and
hockey) were sponsored by publically traded companies.
Sponsoring a sport team is a major decision for a corporation. It is typically a long-term
commitment that requires a significant investment. The average price for acquiring naming rights
of a team’s stadium in the U.S. National Football League in recent years is 120 million dollars for
an average period of 17 years (see details in Table A1 in the appendix). In turn, the sponsoring
company is provided an opportunity to tie the company’s brands with a successful and popular
sport organization, an opportunity to establish a strong relation with a large fan base and the local
community, and a range of effective marketing tools – the sponsor typically gets branding and
signage inside and outside the stadium, product placement rights within the stadium, exclusivity
for use of its products by the team (official sponsor status), and has access to the team’s coaches
and players for promotions.2
1
Extant stadium naming literature focuses on market reaction to the initiation of stadium sponsorship (see, for
example, Clark, Cornwell, and Pruitt (2002), Becker-Olsen (2003), and Leeds, Leeds, and Pistolet (2007)).
2
For example, the NFL team the Dallas Cowboys has recently signed a 25-year, 500 million dollar stadium naming
rights deal with AT&T. As part of the deal, AT&T will continue to invest in improvements to wireless technology at
the stadium, the city of Dallas will get 5 percent of revenue from the deal to help pay off the city's debt, and even the
Legends Drive near the stadium will become AT&T Blvd. (DallasNews.com, July 25, 2013).
3
In light of these potential advantages of sport sponsorship, attaching the company’s name to a
sport team can also be risky. If the team does not perform well, it will likely suffer less exposure
in the media, lower demand for the team’s games and merchandises, and damaged reputation, all
of which can affect the team’s sponsoring company. The nature of competitive sports suggests that
a single event or game can largely determine the success versus failure of the team, and thereby
the value of its sponsoring company.3
In an efficient market therefore post-game stock returns to the sponsoring firms should reflect
rational market reaction to the financial implications of the game outcomes. Yet there may also be
a behavioral effect, which is driven by the unique ability of sport events to generate strong
sentiment at the moment of time among large populations.4 Edmans et al. (2007) argue that
national sport events (especially soccer matches) can produce substantial and correlated mood
swings in a large proportion of a country’s population, which is translated into the stock market
movement in the country. Analyzing publically traded European soccer clubs, Bernile and
Lyandres (2011) conclude that investors are overly optimistic about their teams’ prospects ex ante
and, on average, end up disappointed ex post, leading to negative post-game abnormal returns.
This conclusion is consistent with the study of Brown and Hartzell (2001) on the impact of
basketball game results on the stock price of the NYSE listed Boston Celtics Limited Partnership.
Palomino, Renneboog, and Zhang (2009) find evidence that the abnormal returns for the winning
British soccer teams do not reflect rational expectations but are high due to overreactions induced
3
For example, an article in BusinessWeek observes: “BMW Oracle (ORCL) team sailboat, eliminated in May from
the America's Cup qualifying competition in Valencia, Spain, before the main event even started. German press reports
put the cost of the failed Cup bid at nearly $200 million.” (Jack Ewing, June 7, 2007).
4
Other effects on investor sentiment include sunshine (Saunders (1993) and Hirshleifer and Shumway (2003)),
changes to and from daylight saving hours (Kamstra, Kramer, and Levi (2000)), amount of daylight across seasons
(Kamstra, Kramer, and Levi (2003)), nonsecular holidays (Frieder and Subrahmanyam (2004)), temperature (Cao and
Wei (2005)), lunar cycles (Yuan, Zheng, and Zhu (2006)), and aviation disasters (Kaplanski, and Levy (2010)). See
Hirshleifer (2001) on psychological biases in asset pricing.
4
by investor sentiment. We ask therefore whether sudden change in investor mood as a result of
game outcomes is translated into the market value of the teams’ stadium sponsors. Because the
outcomes of sport events are clear and decisive, we have a reasonable cause to assume that their
effects aggregate across fans/investors, unlike more individual effects, such as prospect theory,
loss aversion, skewness, or habit preferences.
We concentrate on the NFL for several reasons. First, football is the most popular sport in the
U.S. and has been for many years by a wide margin. Football games attract the largest crowds and
achieve the highest television ratings among all major sports.5 Second, a fairly large proportion of
NFL stadiums are sponsored by publically traded companies, 21 out of total of 32 NFL teams (see
details in Table A1). Third, the importance of a single game in the NFL is very high, relative to
the other major sports in the U.S. (baseball, basketball, and hockey). This is because the NFL
season is very short (16 games, compared to at least 82 games in the other sports). Furthermore,
the NFL post-season (playoff) system is based on one game at each progression towards the
championship, compared to a series of games (typically best-of-seven) in the other sports.
The NFL therefore provides a unique setting to assess the impact of outcomes of important and
popular sport matches on the stock price of the sponsoring companies. We manually collected
detailed data on all NFL games for teams with stadiums that are sponsored by publically traded
companies. The sample contains 3,399 games (1,710 home games) during the pre-seasons, regular
seasons, and post-seasons of 21 teams with 26 sponsoring companies, over the period 1997-2013.
5
A 2009 ESPN Sports poll asked respondents to name their favorite spectator sport (defined as one in which the
responder attends games or matches, watches them on TV, listens to them on the radio, or reads about them). The poll
results showed that professional football is the most popular sport with 24.4%, where professional baseball is the
second with 11.0%. The Harris Poll has recently released the results of its annual survey of the favorite sports of
Americans; professional football was the most popular with 35%, followed by professional baseball with 14%
(BusinessInsider.com, January 27, 2014). According to the NFL, during the fall of 2013, NFL games accounted for
34 of the 35 most-watched TV shows among all programming (Bloomberg.com, February 5, 2014).
5
We begin our analysis by looking at home games of NFL teams and comparing the next-day
abnormal stock returns of their stadium sponsors after wins and losses. We measure abnormal
returns using six different models. When all home games are included in the sample, the results do
not indicate any abnormal return after wins or losses. This is not very surprising given that most
games are played simultaneously with other games, are not played in prime time hours, are not
nationally televised, and for the most part their outcome (i.e., the game winner) is expected. We
therefore focus our examination on subgroups of games that attract the highest interest: regular
season games that are played on Monday nights, post-season games, and games with unexpected
outcome. Monday night games and post-season games attract high attention because they are
stand-alone (no other games played at the same time), nationally broadcasted by a major television
network, played at prime time hours (the Monday night games), and are critical for the team
success (the post-season elimination games). An unexpected outcome of a game naturally attracts
more media attention and also carries more “news” to the fans/investors, thus is likely to generate
a stronger post-game stock market reaction. We classify a game outcome of a team as unexpected
using two criteria: if the game outcome is contrary to the pre-game betting spread prediction or if
it comes after a sequence of games with the opposite outcome (i.e., a loss after a series of wins).
The results indicate that games' outcomes affect the market value of the stadium sponsoring
companies. Wins in Monday night games have an economically and statistically significant
positive effect on sponsor stock price, an average of 0.51 percent across all models of abnormal
return, whereas losses have virtually no effect (average of 0.01 percent). The difference however
between the win and loss returns shows relatively weak statistical significance. For post-season
games the difference is greater and somewhat more significant statistically; the losing teams’
sponsors earn average abnormal return lower by 82 basis points than that of the winning teams’
6
sponsors, although the returns are negative both after wins and losses (-0.19 and -1.01 percent,
respectively). This can be explained by the crucial effect a playoff game loss has on the team, as
it eliminates the team from the playoff contention. This may also indicate a surprise effect, i.e., the
home team is expected to win in the post-season. A playoff game is typically played at the stadium
of the team with the better record; thus a win does not have much effect while a loss leads to a
significant negative return of the home team's sponsor.
Games with unexpected outcome generate a similar economic effect to that of playoff games,
but stronger statistically likely due to the larger sample size. The sponsor of the home team earns
on average a positive abnormal return after wins and a negative abnormal return after losses,
yielding a significant win-loss return difference of 81 basis points. This emphasizes the importance
of the element of surprise in moving the stock price of the sponsoring companies.
We confirm the post-game abnormal returns using a pooled regression. We estimate abnormal
returns for all sponsor-days in the sample (game and no-game days) and regress them against
dummy variables indicating a first trading day after a win/loss of the sponsored teams. The
coefficients of the win and loss indicators are typically positive and negative, respectively, where
the differences between the coefficients are fairly similar to the cross-sectional differences in
abnormal returns after wins and losses. Team-specific analysis shows a positive effect for the
majority of the teams, where there is no clear relation between the magnitude of the effect and the
local market size of the team.
We next explore the drivers for the effect of match outcomes on sponsors’ stock returns. As
discussed above, we distinguish between two mechanisms. On the one hand post-game returns to
the sponsoring companies should reflect changes in expected cash flows due to the financial
7
implications of the game outcomes. On the other hand outcomes of sport events are associated
with investor sentiment, which is often reflected in stock price movements.
We assess the presence of investor sentiment using two tests. First, we look at returns on the
sponsors' stock after away games (games played at the opponent's stadium). Rational expectations
suggest a similar effect of wins/losses in away games and home games, as both should have similar
implications for the team’s success (i.e., wins/losses count the same in home and away games) and
thereby the exposure of its sponsor for the remaining of the season. Behavioral bias suggests a
stronger effect for home games; this is because the stadium sponsoring company is only visible
during home games, for which investors may associate the game outcome with the sponsor. The
results generally show substantial reductions in sponsor abnormal returns after away games, which
is consistent with investor sentiment.
Second, the effect of investor sentiment on the stock return in the first trading day after the
game can also be associated with subsequent returns; i.e., an initial overreaction to game outcome
is more likely to reverse in the following days. Our results however are mixed; while in certain
cases the next-day abnormal return tends to reverse, other cases exhibit return continuation.
Considering the results of the two tests, we conclude that investor sentiment plays a partial role in
shaping the market value of the sponsoring firms after important games.
In the final part of the study we investigate whether the sponsors’ post-game stock return
patterns provide profit opportunities. We form a weekly zero-investment portfolio of buying the
stocks of all sponsoring companies whose teams won that week and selling the stocks of all
sponsoring companies whose teams lost that week. We hold this portfolio from the second trading
day to the fifth trading day after the game (as the first-day profit cannot be earned). The portfolio
generates abnormal profits, particularly for home games whose outcomes are hard to predict and
8
by nature attract more attention and provide a higher element of news: mean excess return and
factor-model alphas of approximately 28 percent per NFL season.
The paper contributes to the literature in several aspects. To the best of our knowledge, this is
the first study to highlight the sensitivity of sponsoring corporations’ value to the outcome of
individual sporting events. Although NFL stadium sponsors are typically among the largest and
most well-known firms listed on the U.S. stock exchanges, the outcome of a single match played
by their sponsored teams can lead to significant swings in sponsor firm stock prices. Second, we
provide new evidence of the presence of investor sentiment in stock pricing. Identifying exogenous
factors that can affect the mood among large populations is a key ingredient in studying investor
sentiment. Prior studies use such factors typically at the aggregate level, for example: sunshine
(Saunders (1993) and Hirshleifer and Shumway (2003)), temperature (Cao and Wei (2005)), and
aviation disasters (Kaplanski, and Levy (2010)). We believe that the highly popular NFL games
serve well as an exogenous factor that can generate massive mood swings. Furthermore, NFL
games and stadium sponsorship allow for analysis at the firm level. Our results indicate that stock
market reaction to game outcomes is driven, at least to a certain extent, by investor sentiment. This
finding is consistent with prior studies attributing countrywide market price movement after sport
matches to changes in investor mood (e.g., Edmans et al. (2007)), and also with the documented
emotional reactions that NFL matches can illicit (see White (1989) and Chang et al. (2012)).
Lastly, in a broader asset pricing view, we show that the effect of game outcome on sponsoring
firm value embeds various profitable stock trading opportunities.
The paper proceeds as follows. Section 2 describes the data and game samples, Section 3
examines the effect of home game outcome on sponsoring firm return, Section 4 tests for the
9
presence of investor sentiment, Section 5 offers post-game stock trading opportunities, and Section
6 concludes.
2. Data and game samples
We manually collected data on all NFL games over the years for teams that are sponsored by
publically traded companies. Using official NFL team websites and stadium websites, we
identified teams who are or have engaged in stadium naming rights agreements with publicly
traded firms and obtained key characteristics of the agreements. We then use team websites as well
as secondary sports websites to gather data on game schedules across the sample period, including
game date, location, score, television coverage, and more.6 Based on the game date, we identified
the first subsequent day of stock market activity for the sponsoring firms. Sunshine Forecasts’
database was used to identify historic betting spreads for each game. Our sample contains 3,399
games (1,710 home games) during the pre-season, regular season, and post-season, representing
21 NFL teams and 26 sponsoring companies over the period 1997-2013. Table A1 lists the sample
teams, stadiums, and sponsoring companies. Table A2 shows the game distribution over the
sample period.
We combine the NFL data with CRSP and Compustat to draw accounting variables and stock
return data on the sponsoring companies. Table 1 presents descriptive statistics for the sponsoring
companies and for all CRSP/Compustat firms over the same sample period. Not surprisingly, firms
that sponsor home stadiums of NFL teams are typically much larger than the average firm, have
less growth opportunities (indicated by higher book-to-market ratios), and higher leverage ratios.
6
In addition to NFL.com, secondary websites included Sports Illustrated (sportsillustrated.cnn.com), ESPN
(espn.go.com), and ProFootball Weekly (profootballweekly.com).
10
Stocks of sponsoring firms are highly traded and highly liquid; trading volume, Amihud’s
illiquidity measure, and bid-ask spread are all significantly different than those of the average firm.
Sponsors’ stocks are also less volatile (measured by the standard deviation of daily stock returns
over a month) but with market beta somewhat higher than that of the full sample.
Considering the full sample of home games, which includes all pre-season, regular season, and
post-season games, we do not expect a strong effect. Most NFL games are played simultaneously
with other games, are not played in prime time hours, are not nationally televised, and do not attract
particular post-game media attention unless their outcome is really unexpected. Hence, the stadium
sponsoring companies are not visible to a large, national audience for a typical game. We therefore
focus on several samples of home games that provide very high visibility and are highly important
for the sponsoring companies.
The first sample contains all regular season games that are played on Monday night (92 home
games). Monday night games are distinct. A game held on Monday night is the last game played
in the NFL week (Thursday to Monday), receives exclusive game-day publicity at a national level,
is always played in prime-time hours (typically at 8:30pm EST), and is nationally televised. In
addition, Monday night games are usually chosen based on the importance and the general interest
of the game. For many years ‘Monday Night Football’ has been one of the highest-rated television
shows in the U.S. This means that sponsoring companies are more visible for games held on
Monday night in their stadiums, and thus are more likely to be affected by the outcome of the
games. Important to this study, there is also typically a 24 hour gap between the start time of the
last game on Sunday and the Monday night game. This allows the market reaction measured to be
isolated from the reaction to other NFL-week games.
11
Our second tested sample includes games that are also very visible, but much more important
– the post-season (playoff) elimination games (57 home games).7 As with Monday night games,
post-season games are stand-alone and are typically nationally broadcasted by a major television
network, thus providing high visibility to the stadium’s sponsoring firm. But more importantly,
post-season games are the most meaningful games for the teams, as their outcome solely
determines if the team will continue to compete for the championship (in case it wins) or will be
eliminated from the competition (if it losses). The perception of a successful season versus a failed
one is often determined by a single post-season game. The outcome of post-season games therefore
can have a direct impact on the visibility of the sponsors in the rest of the season.
Our third and final sample consists of all games whose outcome is unexpected. These games
are likely to get more media attention after the game, making the stadium sponsoring companies
more visible. In addition, if a game win/loss is unexpected, it creates news of larger magnitude for
the team and thereby for its sponsor. An unexpected win or loss can be viewed as any other
corporate news that carries value for the company, and thus can have stronger stock market
reaction. To determine whether a game outcome is expected or unexpected we use information
obtained from pre-game betting spreads and prior team performance. A game outcome is
considered expected if it is consistent with the sign of the betting spread or if it comes after a
sequence of at least three games with the same outcome. In the same way, a game outcome is
classified as unexpected if it is against the sign of the betting spread or if it comes after a sequence
of at least three games with the opposite outcome. To eliminate trivial game predictions, we
consider only spreads of at least 5 points. Game outcomes for which the two categories of betting
spread and prior win/loss runs conflict are not considered as either expected or unexpected. And
7
The post-season sample does not include the Super Bowl game as it is played at a neutral stadium.
12
games with spreads lower than 5 points and without prior runs are considered as unpredictable ex
ante. Our sample of interest, games with unexpected outcome, contains 278 home games.
Table A2 shows a fairly uniform distribution of the samples’ games over the years, which
provides a solid ground for our examination. That is, the results are not likely influenced by an
unusual effect in a specific season, or by cross-sectional dependence induced by same-day
clustering (see Brown and Warner (1985)). This is verified by robustness tests accounting for timeclustering.
3. Effect of home game outcome on sponsoring firm return
We investigate the effect of game outcome by estimating the abnormal return to the sponsoring
companies in the first trading day after each game played in their stadium and comparing the
abnormal returns after wins and losses. We verify the robustness of the results using pooled
regression, wherein we calculate abnormal returns for all sponsor-days in the sample and regress
them against variables indicating first-trading days after home game wins/losses. Details and
results are below.
3.1 Calculating abnormal return
We begin our empirical analysis by estimating abnormal stock returns of the stadium
sponsoring companies in the first trading day after the game (referred to as ‘post-game day’). The
firm’s abnormal stock return is the difference between its raw return and its expected return for
that day. To mitigate the sensitivity of the results to a specific model of expected return, we employ
six different models that are commonly used in the literature (for detailed analyses of the models
see Brown and Warner (1985) and Barber and Lyon (1997)). The first model is the mean-adjusted
model: expected return is estimated by averaging the firm’s raw returns during the past 250 trading
13
days prior to the game. The second is market-adjusted model: expected return is estimated by the
value-weighted market index on the post-game day. The third is market model: expected return is
estimated by the fitted value of the stock return on the post-game day, based on a regression of the
sponsoring firm’s raw return on the value-weighted market index return during the past 250 days
prior to the game. The fourth is factor model: expected return is estimated in a similar way to the
market model with a larger set of explanatory variables, including the Fama and French (1993)
three factors, the momentum factor (all factor returns are downloaded from Kenneth French’s
website), dummy variables indicating January and Monday, and lagged return on the sponsor’s
stock. The fifth is reference portfolio: expected return is estimated by the equal-weighted average
return of firms in a size/book-to-market portfolio that includes the sponsoring firm; portfolios are
formed by first sorting all stocks into ten equal deciles according to the firm’s size as of the
beginning of the post-game day, and then within each decile, sorting all stocks into five equal
book-to-market quintiles. The sixth is matched (control) firm: expected return is measured by the
return of the firm with the closest book-to-market ratio within the same size decile as the
sponsoring firm.
3.2 Post-game day abnormal return
In Table 2 we look first at the effect of “typical” NFL games, i.e., games that are comparably
not highly visible or important at the onset, and that do not produce unexpected outcomes that
attract media attention. We present the abnormal returns according to the six models described
above for the game samples, as well as the results based on the average across all models. The tstatistic of the average abnormal return is thus calculated using the cross-sectional properties of
14
the average, which should capture the dependency structure across the individual models.8 All
returns are reported in percent.
As expected, when all home games are included in the sample, the results do not indicate any
abnormal return after wins or losses. In fact, the average abnormal return after a loss of the home
team is higher by 10 basis points than that after a win, yet is not statistically significant. Outcomes
of unpredictable games also do not lead to any significant effect on the sponsors’ stock price, but
interestingly, expected outcomes lead to a negative effect: a win-loss return difference of -0.60
percent with a t-statistic of -3.07.
The negative effect of expected outcomes may seem puzzling, i.e., why would expected wins
lead to negative returns and expected losses to positive returns? A possible explanation for this
result can be related to fans’ behavior prior to sport matches. Several studies argue that fans often
tend to overreact to the expected performance of their teams in certain events (see, e.g., Krizan
and Windschitl (2007) and Bernile and Lyandres (2011)). This means that fans can be overly
optimistic for a game that their team expects to win and overly pessimistic if the team expects to
lose. Because a game with an expected outcome should not have a material effect on the sponsor
stock price, a post-game correction to pre-game overreaction will lead by itself to a negative effect
of the game outcome.
Turning our focus to highly visible and important games, the results show that game outcome
has a positive effect on sponsor stock price (reported in Table 3). After Monday night games, the
home team’s sponsor earns a positive abnormal return if the team had won, an average of 0.51
8
For robustness we also applied Bayesian model averaging on the six models (see Avramov (2002)). Assuming a
prior uniform distribution of the validity of the different models, we regressed daily raw returns on the models’
benchmarks of expected return, calculated the posterior probabilities that each benchmark is the correct one given the
data, and used these posterior probabilities as weights for averaging the models’ abnormal returns. The Bayesian
averaging yields almost identical results as the equal-weighted averaging (not reported).
15
percent with a t-statistic of 2.18, and no abnormal return if the team had lost (average of 0.01
percent). Although the average difference between the win and loss post-game day abnormal
returns is large in economic terms, its t-statistic is only 1.40. This is partially due to the relatively
small sample of Monday night games (92 home games).
For post-season games the impact is larger economically and somewhat more significant
statistically. Averaging across all models, the losing teams’ sponsors earn an abnormal return
lower by 0.82 percent than that of the winning teams’ sponsors (t-statistic of 1.72). All six
abnormal return models yield meaningful effects (differences in win-loss abnormal returns of 58
to 127 basis points). Unlike the Monday night games, the effect of playoff games seems driven by
negative returns, as the average post-game day abnormal return is negative both after wins and
losses (-0.19 and -1.01 percent, respectively). This can be explained by the crucial effect a playoff
game loss has on the team. A playoff loss eliminates the team from the playoff contention,
effectively ending the team’s season. A playoff win only guarantees to keep the team in the
competition for one more game. While wins and losses in the regular season serve to seed the team
for playoff matchups, wins and losses in the post-season ultimately determine the success or failure
of the team’s entire season. The post-season results may also indicate that there is a surprise effect.
As a playoff game is typically played at the stadium of the team with the better record, there is a
greater expectation that the home team will win in a post-season game; thus a win does not have
much effect while a loss leads to a significant negative return to the home team's sponsor.9
Games with unexpected outcome show a similar impact to that of the post-season games, but
results are more significant statistically, likely due to the larger sample size. The sponsor of the
9
Note that although post-season games are played during the weekend, their post-game day negative abnormal returns
are not driven by the weekend effect (see, French (1980)). This is because for most models of abnormal return, the
expected returns are estimated from Monday returns as well.
16
home team earns on average a positive abnormal return if the team had won and a negative return
after a loss, yielding a win-loss post-game day return difference of 0.81 percent with a t-statistic
of 2.50. This result indicates the importance of the element of surprise in moving the stock price
of the sponsoring company.
The mean differences in abnormal returns after wins and losses reported in Table 3 are
meaningful. Clark et al. (2002) find that the sponsorship agreement announcement increases the
sponsor’s stock price by 1.65 percent on average in the four major sports in the U.S. We show that
the outcome of a single game affects the sponsor’s stock by an average return of 0.50 to 0.82 across
the samples of Monday night, post-season, and unexpected outcome games. These effects are also
comparable and even stronger than those reported in Edmans et al. (2007). They find that at the
national level a loss in the soccer World Cup elimination stage leads to a next-day abnormal stock
return of -0.49 percent, whereas a win does not lead to a significant positive return.
3.3 Team analysis and market size
An interesting aspect of the potential impact of game outcomes is the size of the media market
of the team. An argument could be made for a stronger effect in various markets. The effect could
be stronger in smaller markets, where the fans are generational, teams are more involved in the
community, and media coverage is more likely to be supportive. Yet areas with a lower population
density may have a weaker influence on the stock market and generally have fewer media channels
to report on team activities and mention stadium sponsors. The effect could be stronger in larger
markets. These markets have a greater population density and more media outlets/reporters. Yet
large markets usually have multiple sports teams, creating a dilutive effect to game and media
coverage.
17
To examine the effect of the local market size, we look separately at each individual team.
Because many teams have only a few observations, we combine for each team the three groups of
Monday night games, post-season games, and games with unexpected outcome, and we include
only teams with at least five wins and five losses across these groups. This provides us with a total
of 15 teams. Figure 1 shows the team-specific mean difference in the sponsoring companies’
abnormal stock returns after home game wins and losses, averaged across the six models of
abnormal return. The teams are ordered by the size of their media market, measured by population
in TV households within a 75-mile radius of the team’s stadium.10
The figure indicates that home game outcome has a positive effect on sponsor stock price for
the majority of the teams, reinforcing the general results: nine teams show a large positive effect
(win-loss returns between 0.6 and 2.0 percent); three teams show a positive, but low effect; two
teams show a low negative effect; and one team shows a negative effect of -0.53 percent. The
impact of the game outcome however is not clearly related with the local market size of the team.
Teams located in the middle of the scale seem to have relatively small effects, and the two teams
with the largest markets in the sample (New England Patriots and Oakland Raiders) do not exhibit
any significant effect. This is not surprising given the aforementioned opposing effects of market
size.
As a final note, we acknowledge that New York is a special case. Both the Giants and the Jets
play in MetLife Stadium, which is close to a market epicenter and a high population density. In
addition, the stock market is located in New York thus local sports fans may be more involved in
stock trading. The New York market and MetLife Stadium also have the highest exposure to
dilution as eight major sports teams claim affiliation with the city. Unfortunately, the MetLife
10
The NFL defines a team's "local" market as all the TV markets that lie within a 75-mile radius of the stadium.
18
sponsorship started in August of 2011, providing us with too few observations for a city-specific
analysis at this time. Perhaps in the future, these data could provide a unique opportunity to
examine investment behavior of regionally specific populations.
3.4 Robustness test: pooled regression
To confirm the impact of game outcomes on stadium sponsors’ stock returns documented
above we employ a pooled time-series regression (for similar specifications see Edmans et al.
(2007) and Bernile and Lyandres (2011)). Using the six models of expected return described
above, we estimate the abnormal return for each of the sponsor-days in the sample. For the full
sample of sponsor-days (game and no-game days), we run a pooled regression of the six-model
average abnormal return on two dummy variables: Win indicates a first trading day after a home
game win of the sponsored team, and Loss is the equivalent indicator for a home loss of the
sponsored team. The loadings on these dummy variables thus capture the marginal effects of win
and loss on the sponsoring company’s stock price.
This regression offers the advantage of exploiting the full time-series of the sponsors’ stock
returns, it allows estimating the effects of wins and losses independently, and it controls for
potential systematic biases in the estimates of expected returns. For example, if a specific sponsor
is relatively small in its reference portfolio, then a positive abnormal return to this sponsor’s stock
after a game win can be partially attributed to its relatively small size. Comparing the sponsor’s
post-game day abnormal return to its time series of abnormal returns that are estimated in the same
way should eliminate this size effect.
The regression model also allows us to address a potential effect of cross-sectional dependence.
Our samples generally show a lower level of time-clustering. The playoff sample consists of 57
games that are played over 42 different weekends, where no more than three games are held in any
19
of the weekends. The sample of Monday night games has almost no clustering because there is
usually a single Monday night game every week, and the sample of games with unexpected
outcome is not highly clustered as well (278 games played over 184 different weekends). Yet, as
prior studies show that even relatively small cross-correlation can lead to over-rejection of the null
hypothesis (see e.g., Kolari and Pynnönen (2010)), we examine the robustness of the results to
game-day clustering. In addition to the OLS regression we estimate a Generalized Least Squares
(GLS) regression that accounts for group-wise clustering by trading day.
The OLS regression estimates reported in Panel A of Table 4 are consistent with the abnormal
return averages in Table 3. As the results above, Monday night games show a strong win effect
and no significant loss effect (the t-statistics of the win and loss coefficients are 2.37 and 0.08,
respectively), whereas post-season game effect is driven solely by home team losses (t-statistics
of win and loss of -0.54 and -2.26, respectively). For games with unexpected outcome, a win has
a strong positive effect (t-statistic of 3.03), where a loss has a negative effect but not as significant
(t-statistics of -1.22).
Moreover, across the three samples, the differences between the coefficients of the win and
loss dummy variables, which represent the marginal influence of the game outcome on the
sponsor's stock return, are very similar to the win-loss average abnormal return differences
reported in Table 3; difference in coefficients of 0.50 to 0.84 percent (with p-values of 0.06 and
lower), compared with win-loss abnormal return difference of 0.50 to 0.82 percent. Finally, the
GLS regressions reported in Panel B show similar and even slightly more significant results than
the OLS regressions, confirming that the t-statistics are not inflated by time-clustering.
The results in this section consistently indicate that the market values of NFL sponsors are
affected directly by the outcomes of meaningful games played in their stadiums.
20
4. Testing for investor sentiment
The findings in Section 3 raise the question regarding the mechanism driving the impact of home
game outcomes on the sponsoring firms’ stock price. On the one hand, outcomes of important
games can have real financial implications for the sponsoring companies. A team that just lost an
important regular season game that affects its chances to compete for the championship in the
current season will naturally attract less attention in the remainder of the season. And a team that
just lost a post-season elimination game has also ended the NFL-affiliated media mentions of their
stadium, effectively ending the sponsor’s naming rights campaign until the next game is played in
the stadium. These losses mean reduced media coverage, TV ratings, home game attendance, and
demand for team products – all of which should affect the sponsoring firm’s future cash flows. A
reverse effect is expected had the firm won the game. A rational expectations argument suggests
therefore that the post-game day change in the sponsor’s stock price reflects changes in expected
cash flows.
On the other hand, outcomes of sport events are correlated with sudden change in investor
mood, which is often reflected in stock price movements (see Boyle and Walter (2002) and Edmans
et al. (2007)). This argument can apply specifically to the NFL due to its high popularity and the
strong emotions it generates. For example, Chang et al. (2012) find that game outcome of NFL
teams affects the value of locally headquartered stocks, where White (1989) documents that
elimination from NFL playoff games leads to a significant increase in homicides in the cities
following the games. The question is whether the ability of NFL games to generate strong
sentiment among the teams’ fans lead to stock market overreaction for the sponsoring companies
attached to the teams.
21
Note that we address the behavior of the stock investor and not that of the consumer. One can
argue that outcomes of NFL games can create consumer sentiment, resulting in over-consumption
of the products of the sponsoring companies. However, if the sponsor stock price increases because
investors expect over-consumption, this would qualify as rational stock trading behavior because
this consumption builds real value to the sponsor.
The results so far can provide some indication for whether the post-game day abnormal stock
return to the sponsoring firm is driven by rational expectations or investor sentiment. For example,
the evidence that Monday night home games affect sponsors’ return, but not necessarily any
regular season home game, is more consistent with the investor sentiment argument. This is
because the main difference between Monday night games and all other regular season games is
visibility, not the level of importance. That is, if two games are equally important for their teams,
they should create the same impact on the sponsoring firms’ value. But if the more visible game
generates a stronger impact, it is likely driven by correlated change in fan sentiment towards the
team and thereby its sponsor. In contrast, the evidence that outcomes of post-season elimination
games create stronger impact than those of Monday night games is consistent with the rational
expectations argument. This is because both types of games are very visible, but the post-season
games are, on average, much more important than Monday night games (and any other regular
season games).
To further explore the presence of investor sentiment, we look at whether there is an effect
also in away games (games played at the opponent's stadium) and whether the next-day sponsor
return tends to reverse in the following days.
22
4.1 Effect of away games
Rational expectations suggest a similar effect of wins/losses in away games and home games,
as both should have similar implications for the team and the sponsor for the remaining of the
season (i.e., wins/losses have the same count in home and away games). Behavioral bias suggests
a stronger effect for home games than for away games; this is because the stadium sponsoring
company is only visible during home games, and thus investors may associate the game outcome
with the sponsor.
We replicate the main results of the prior tables for away games, and summarize the
comparison of home/away games in Figure 2. Both mean difference and pooled regression results
indicate that Monday night games and games with unexpected outcome that are played away lead
to much weaker and insignificant abnormal return on the sponsors’ stocks; an effect of 19 to 22
basis points after away games, compared to 50 to 84 basis points after home games. These results
strongly support the role of investor sentiment.
In the post-season, however, the effect of away games is more similar to that of home games.
The average difference in abnormal returns between wins and losses is actually higher in away
games than in home games (0.87 percent compared with 0.82 percent). The difference between the
pooled regression coefficients of the win and loss indicators is reduced from 0.82 to 0.52 percent
in away games. This may suggest that the effect of playoff games on the teams’ sponsors is driven
mostly by rational expectations and only to a minor degree by investor sentiment.
Figure 2 captures much of the message of the findings so far: outcomes of important home
games affect the market value of the team stadiums’ sponsors; away games generate a significantly
weaker effect for the most part, leading to the conclusion that abnormal returns to the sponsors’
stocks are driven at least to a certain extent by investor sentiment.
23
4.2 Subsequent return reversal
We examine whether the post-game day abnormal return tends to reverse in the following days,
as evidence of initial overreaction to game outcome. We first estimate the cumulative average
abnormal return during the four days following the first post-game trading day (days 2 to 5). We
choose a window of four days as it gives sufficient time for return reversal, yet is not affected by
the next game outcome (NFL games are typically played once a week). In addition, we estimate a
cross-sectional regression of the cumulative abnormal return in days 2 to 5 on the first trading day
abnormal return. A negative slope coefficient would suggest a return reversal, where a slope
coefficient equals to -1 would indicate a complete reversal.
The results in Panel A of Table 5 do not indicate any evidence for return reversal. For playoff
games, the difference between abnormal returns after wins and losses continues to be positive in
the subsequent days; a cumulative abnormal return of 3.0 percent in days 2 to 5 with a t-statistics
of 1.28. For the other samples the cumulative win-loss abnormal return difference in the
subsequent days is very low and insignificant.
The cross-sectional regression in Panel B however yields mixed results. An indication for
partial return reversal appears for games with unexpected outcome, where the slope coefficient is
negative and significant for the sample of game wins. The post-season games show weak evidence
for return reversal after wins, yet a strong evidence for return continuation after losses. After
Monday night games the results indicate mainly return continuation. While these results do not
support an initial overreaction for the most part, we cannot rule out the presence of behavioral
biases, as almost all cases exhibit significant relations (positive or negative) between the next-day
and subsequent days returns.
24
Considering the results of the two tests distinguishing between rational market reaction and
investor sentiment (home/away games and subsequent returns), we can only conclude that investor
sentiment plays a partial role in shaping the market value of the sponsoring firms after important
games.
Attempts to conduct additional tests to explore the presence of investor sentiment do not yield
definitive results. For example, prior studies have argued that investor sentiment is correlated with
trading characteristics, including the timing and volume of trading, the type of traders (institutional
vs. individual), the extent of order imbalance, and others. Building on these studies, we examine
whether the effect of home game outcome on the stadium sponsor market value is associated with
unusual levels of several trading variables. Our results provide only marginal evidence of
behavioral trading after home games (not reported).
4.3 Discussion
Investor sentiment has been discussed theoretically and explored empirically in the literature
mostly at the aggregate or market level; Saunders (1993), Hirshleifer and Shumway (2003),
Kamstra, Kramer, and Levi (2000, 2003), Baker and Wurgler (2007), Edmans et al. (2007) and
Kaplanski and Levy (2010) to name only a few. We believe that extending marketwide or
countrywide analyses such as in Edmans et al. (2007) to the firm-level can enhance our
understanding of investor sentiment.
First, analyzing cross-sectional variation in sentiment among stocks traded in the same market
eliminates marketwide or countrywide sensitivity to sentiment. For instance, stock markets around
the world are likely to exhibit different levels of market efficiency, and thereby have different
exposures to investor sentiment. Moreover, sentiment can affect individual stocks in different ways
(see, e.g., Baker and Wurgler (2006)).
25
In addition, and perhaps most importantly, we explore the presence of investor sentiment in a
unique setting that puts together two extreme sides of the scale of sentiment. On the one hand there
are firms that are least likely to be affected by behavioral biases, namely the sponsoring companies.
These companies are among the largest companies traded in the U.S. stock market, they are highly
liquid, highly traded, highly visible, and are less affected by market frictions (e.g., Baker and
Wurgler (2006) suggest that sentiment is more likely to affect small and young companies with
relatively low levels of asset tangibility and profitability). On the other hand, due to sport stadium
sponsorship agreements, these firms are directly linked to frequent events that generate strong
emotions at the moment of time among large populations, especially the highly popular NFL
games. Having this contrast, our findings indicate that investor mood and emotions do affect the
market values of the sponsoring firms. We show therefore that investor sentiment is not limited
only to small and less popular stocks, it can also affect the largest and most transparent stocks on
the exchange.
It is important to acknowledge however that although this study presents new aspects of
investor sentiment, the magnitude of the impact of sentiment as well as the mechanisms triggering
sentiment, especially across firms, are far from being completely understood. For example, while
post-game trading behavior and the reduced effect in away games indicate behavioral biases, the
patterns of subsequent returns are quite puzzling. Specifically, next-day abnormal return tends to
reverse for certain samples, while other samples exhibit return continuation. This can raise several
questions; for example, does return continuation reflect initial underreaction to true value
generated by the game outcome, or continued overreaction driven by change in investor mood?
More generally, is the impact of investor sentiment on stock prices as a result of isolated events
limited to one or only a few trading days, or reflected in longer horizons? We do not provide
26
answers to these questions in this study. We do however explore in the next section whether these
unusual patterns in subsequent returns embed profitable trading opportunities, which are often
used as evidence for stock mispricing/behavioral bias.
5. Post-game trading opportunities
The findings above suggest that the market reaction to game outcome is not limited to the next
trading day, but rather seems, at least for some of the samples, to continue in the following days.
Assuming that the next-day return cannot be realized by investors, we examine whether one can
earn abnormal profits by trading sponsors’ stocks in the subsequent days.
Because profitable trading rules require a sufficiently large number of traded securities at the
moment of time, we do not limit our examination to the game samples analyzed in the previous
section as they cannot provide more than one or only a few games in a given week. We rather
consider two large groups of games. The first group contains all home games, capturing the general
impact of games played in sponsored stadiums. The second group contains the home games whose
outcome is most unpredictable ex ante. The motivation for this distinction is that any outcome in
unpredictable games, win or loss, provides some element of news that can lead to sharper stock
market reaction. We therefore include in this group only games with very low betting spreads, 3
points and lower in absolute value.11
We examine the performance of the following long-short investment strategy. Every week
during the NFL season we form a value-weighted portfolio of buying the stocks of all sponsoring
companies whose teams won that week and selling the stocks of all sponsoring companies whose
teams lost that week. We hold this portfolio from the second trading day after the game until the
11
A cutoff of 3 is natural as it represents the points awarded for a field goal in an NFL game.
27
fifth trading day. Table 6 shows the portfolios’ mean excess weekly returns (in excess of the riskfree rate) and alphas from factor models over the years 1997-2013. The CAPM one-factor model
uses the market factor. The three factors in the 3-factor model are the Fama and French (1993)
factors. The 4- and 5-factor models include the Fama-French factors augmented with momentum
and reversal factors. All returns and alphas are in percent per week.
When applied to all home games, the portfolio shows some evidence of abnormal profit; the
mean weekly excess return is around 0.4 percent, and factor-model alphas are slightly higher,
where the t-statistics are 1.70 to 1.83. The results are very strong when the investment is applied
to most unpredictable home games. The mean excess return and factor-model alphas are 1.45 to
1.52 percent per week with t-statistics between 3.23 and 3.34. Considering that a regular NFL
season runs for 17 weeks, this investment strategy yields an abnormal profit of approximately 28
percent per season. Abnormal returns following home games, especially unpredictable ones, can
thus be translated into profitable trading opportunities.
To verify the importance of game outcomes for the success of NFL sponsor trading, we
examine if one can achieve abnormal profits by trading sponsors’ stocks prior to the game. Every
Friday during the NFL season we form a value-weighted portfolio of buying the stocks of all
sponsoring companies whose teams expect to win over the weekend and selling the stocks of all
sponsoring companies whose teams expect to lose. We hold the portfolio until the next Friday.
This pre-game trading strategy however does not generate positive profits (results are not
reported), as opposed to the post-game strategy.
The abnormal profit generated by the winning-losing trading strategy is informative in several
aspects. First, the magnitude of the profit is comparable and even higher than those generated by
the most puzzling return patterns, such as momentum (see Jegadeesh and Titman (1993)) and
28
distress effect (see Dichev (1998); Campbell, Hilscher, and Szilagyi (2008)). Second, while we do
not have a convincing explanation for this abnormal profit, we acknowledge that it can be driven
by behavioral bias. This can provide further indication for the presence of investor sentiment in
the sponsors’ stock price movements.
Lastly, an important aspect of the post-game effect is that it involves only very large
companies, the teams’ stadium sponsors. These companies are typically highly traded, highly
liquid, have less information asymmetry, and are strong financially. This means that the portfolios’
profits document above are not likely driven by market frictions and are easier to implement,
compared to other pricing effects. For example, Griffin and Lemmon (2002) document that the
value anomaly is stronger among distressed stocks; Avramov et al. (2007) point to momentum
being present mostly among low credit rating stocks; and Eisdorfer (2008) shows that
approximately 40 percent of the momentum profit is generated by delisting returns. The NFL
teams’ stadium sponsors are not likely to default or be delisted from the stock exchange. In
addition, the profits generated by anomalous stock return portfolios typically require massive short
selling, which is not always feasible for many stocks (e.g., low liquid stocks, distressed stocks). In
that respect, therefore, it is fairly easy to form the long-short NFL sponsors’ portfolios.
6. Conclusions
We document that stock prices of companies that sponsor NFL stadiums are affected by the
outcomes of important games played in the stadiums. Employing six different models of abnormal
return shows that the mean difference between next-day return after a win and after a loss of the
home team is 50 to 82 basis points. Pooled regressions exploiting the full time-series of the
sponsors’ stock returns yield very similar estimates.
29
We explore whether the post-game day abnormal returns represent rational changes in
expectations of future cash flows, or overreaction by investors associating the team’s performance
with its sponsoring company. We find that the effect of game outcomes on sponsors’ stock prices
is significantly reduced after away games, which is consistent with investor sentiment. We do not
find however a clear evidence of return reversal that is assumed to correct for initial overreaction.
We conclude that the impact of home game outcomes is partially driven by investor sentiment.
The post-game return patterns provide profit opportunities. We form a weekly zero-investment
portfolio of buying the stocks of all sponsoring companies whose teams won that week and selling
all sponsoring companies whose teams lost that week. Applying this investment strategy to home
games considered as unpredictable ex ante (using pre-game betting spread data) generates mean
excess return and factor-model alphas of approximately 28 percent per NFL season.
30
References
Ashton, J., B. Gerrard, and R. Hudson, 2003, “Economic Impact of National Sporting Success:
Evidence from the London Stock Exchange,” Applied Economic Letters 10, 783-785.
Avramov, D., 2002, “Stock Return Predictability and Model Uncertainty,” Journal of Financial
Economics 64, 423-458.
Avramov, D., T. Chordia, G. Jostova, and A. Philipov, 2007, “Momentum and Credit Rating,”
Journal of Finance 62, 2503-2520.
Baker, M., and J. Wurgler, 2006, “Investor Sentiment and the Cross-Section of Stock Returns,”
Journal of Finance 61, 1645-1680.
Baker, M., and J. Wurgler, 2007, “Investor Sentiment in the Stock Market,” Journal of Economic
Perspectives 21, 129-151.
Barber, B.M., and J.D. Lyon, 1997, “Detecting Long-Run Abnormal Stock Returns: The Empirical
Power and Specification of Test Statistics,” Journal of Financial Economics 43, 341-372.
Becker-Olsen, K., 2003, “Questioning the Name Game: An Event Study Analysis of Stadium
Naming Rights Sponsorship Announcements,” International Journal of Sports marketing &
Sponsorship 5, 181-192.
Bernile G., and E. Lyandres, 2011, “Understanding Investor Sentiment: The Case of Soccer,”
Financial Management 40, 357-380.
Boyle, G., and B. Walter, 2002, “Reflected Glory and Failure: International Sporting Success and
the Stock Market,” Applied Financial Economics 13, 225-235.
Brown, G.W., and J.C. Hartzell, 2001, “Market Reaction to Public Information: The Atypical Case
of the Boston Celtics," Journal of Financial Economics 60, 333-370.
Brown, S., and J. Warner, 1985, “Using Daily Stock Returns: The Case of Event Studies,” Journal
of Financial Economics 14, 3-31.
Campbell, J.Y., J. Hilscher, and J. Szilagyi, 2008, “In Search of Distress Risk,” Journal of Finance
63, 2899-2939.
Cao, M., and J. Wei, 2005, “Stock Market Returns: A Note on Temperature Anomaly,” Journal of
Banking and Finance 29, 1559-1573.
Chang, S.C., S.S. Chen, R.K. Chou, and Y.H. Lin, 2012, “Local Sports Sentiment and Returns of
Locally Headquartered Stocks: A Firm-Level Analysis,” Journal of Empirical Finance 19,
309-318.
Clark, J.M., T.B. Cornwell, and S.W. Pruitt, 2002, “Corporate Stadium Sponsorships, Signaling
Theory, Agency Conflicts, and Shareholder Wealth,” Journal of Advertising Research 42, 1633.
Dichev, I., 1998, “Is the Risk of Bankruptcy a Systematic Risk?” Journal of Finance 53, 11411148.
Edmans, A., D. Garc'ia, and O. Norli, 2007, “Sports Sentiment and Stock Returns,” Journal of
Finance 62, 1967-97.
31
Eisdorfer, A., 2008, “Delisted Firms and Momentum Profits,” Journal of Financial Markets 11,
160-179.
Fama, E.F., and K.R. French, 1993, “Common Risk Factors in the Returns on Stocks and Bonds,”
Journal of Financial Economics 33, 3-56.
French, K.R., 1980, “Stock Returns and the Weekend Effect,” Journal of Financial Economics 8,
55-69.
Frieder, L., and A. Subrahmanyam, 2004, “Nonsecular Regularities in Returns and Volume,”
Financial Analysts Journal 60, 29-34.
Griffin, J.M., and M.L. Lemmon, 2002, “Book-to-Market Equity, Distress Risk, and Stock
Returns,” Journal of Finance 57, 2317–2336.
Hirshleifer, D., 2001, “Investor Psychology and Asset Pricing,” Journal of Finance 91, 342-346.
Hirshleifer, D., and T. Shumway, 2003, “Good Day Sunshine: Stock Returns and the Weather,”
Journal of Finance 58, 1009-1032.
Jegadeesh, N., and S. Titman, 1993, “Returns to Buying Winners and Selling Losers: Implications
for Stock Market Efficiency,” Journal of Finance 48, 65–91.
Kamstra, M.J., L.A. Kramer, and M.D. Levi, 2000, “Losing Sleep at the Market: The Daylight
Saving Anomaly,” American Economic Review 12, 1000-1005.
Kamstra, M.J., L.A. Kramer, and M.D. Levi, 2003, “Winter Blues: A SAD Stock Market Cycle,”
American Economic Review 93, 324-343.
Kaplanski, G., and H. Levy, 2010, “Sentiment and Stock Prices: The Case of Aviation Disaster,”
Journal of Financial Economics 95, 174-201.
Kolari, J.W., and S. Pynnönen, 2010, “Event Study Testing with Cross-Sectional Correlation of
Abnormal Returns,” Review of Financial Studies 23, 3996-4025.
Krizan, Z., and P.D. Windschitl, 2007, “Team Allegiance can lead to both Optimistic and
Pessimistic Predictions,” Journal of Experimental Social Psychology 43, 327-333.
Leeds, E.M., M.A. Leeds, and I. Pistolet, 2007, “A Stadium by Any Other Name: The Value of
Naming Rights, Journal of Sports Economics 8, 581-595.
Palomino, F., L. Renneboog, and C. Zhang, 2009, “Information Salience, Investor Sentiment, and
Stock Returns: The case of British Soccer Betting,” Journal of Corporate Finance 15, 368387.
Saunders, E.M., 1993, “Stock Prices and Wall Street Weather,” American Economic Review 83,
1337-1345.
Scholtens, B., and W. Peenstra, 2009, “Scoring on the Stock Exchange? The Effect of Football
Matches on Stock Market Returns: An Event Study,” Applied Economics 41, 3231-3237.
White, G., 1989, “Media and Violence: The Case of Professional Football Championship Games,”
Aggressive Behavior 15, 423-433.
Yuan, K., L. Zheng, and Q. Zhu, 2006, “Are Investors Moonstruck? Lunar Phases and Stock
Returns,” Journal of Empirical Finance 13, 1-23.
32
Table 1. Descriptive statistics
The table presents descriptive statistics for the sample of NFL stadium sponsoring firms and for all
CRSP/Compustat firms over the period 1997-2013. For all variables, observations outside the top and the
bottom percentiles are excluded. Size is market equity value (in millions of dollars). Book-to-market is
book equity value divided by market equity value. Leverage is the ratio of book value of total debt to book
value of total assets. Daily stock return is the daily return over the sample period, and Stdev of stock returns
is the standard deviation of the daily stock returns in a calendar month (both are reported in percent). Market
beta is measured by regression of stock return on market return over the past 60 months. Trading volume
is the monthly volume over the sample period (in thousands). Amihud’s illiquidity is the monthly average
of daily ratios of absolute return to dollar trading volume (in millions). Bid-ask spread is the difference
between the stock’s closing ask price and bid price, divided by the average bid and ask prices (reported in
percent).
Interpretation: Firms that sponsor stadiums of NFL teams are larger, have less growth opportunities, and
are more leveraged than the average firm. Sponsors’ stocks are also highly liquid and highly traded, and
are less volatile.
NFL sponsors (26 firms)
Mean
Median
Stdev
Size
Full sample (18,426 firms)
Mean
Median Stdev
27,740.6
12,919.7
40,677.8
1,786.2
207.9
5,485.3
Book-to-market ratio
0.667
0.417
1.028
1.023
0.541
1.497
Leverage ratio
0.295
0.311
0.197
0.217
0.166
0.214
Daily stock return
0.032
0.000
2.508
-0.120
0.000
3.781
Stdev of stock returns
2.115
1.775
1.268
2.949
2.438
2.021
Market beta
1.151
0.988
0.757
1.036
0.905
0.895
138.27
38.46
389.41
10.80
1.45
29.10
Amihud’s illiquidity
0.049
0.004
0.185
11.475
0.320
43.722
Bid-ask spread
0.430
0.116
0.712
2.101
0.714
3.457
Trading volume
33
Table 2. Post-home game day abnormal returns: Typical games
The table presents the average abnormal stock returns of NFL stadium sponsoring companies in the first
trading day after a home game of their sponsored teams over the period 1997-2013. Abnormal returns are
presented for all home games and separately for wins and for losses of the home teams, within three
samples. The first sample contains all games (including pre-season, regular season, and post-season games).
The second sample contains all games with expected outcome; a game outcome is classified as expected if
it either meets the prediction of a pre-game betting spread of at least 5 points or comes after at least three
consecutive games with the same outcome. The third sample contains all games whose outcomes are
unpredictable ex ante; these includes games with betting spreads smaller than 5 points and with no prior
runs of at least three wins or losses. Abnormal return is the difference between the raw return and the
expected return, as measured by the six models described in Section 3.1. The table also reports the results
based on the average of all models. The t-statistic of the average abnormal return is calculated using the
cross-sectional properties of the average. All returns are in percent and t-statistics are in parentheses.
Interpretation: For NFL home games that are not highly visible or that do not produce unexpected
outcomes, wins (losses) do not increase (decrease) the market value of the sponsoring companies. The
samples of all games and unpredictable games do not exhibit any significant return after wins or losses,
whereas games with expected outcome generate a negative effect of 60 basis points (t-statistic of -3.07).
Model of abnormal return
Market
Factor Reference Matched Average
model
model
portfolio
firm
Abn return
Meanadjusted
Marketadjusted
1,019
-0.014
(-0.16)
-0.029
(-0.39)
-0.042
(-0.63)
-0.041
(-0.74)
-0.005
(-0.08)
-0.028
(-0.29)
-0.027
(-0.40)
691
0.108
(0.85)
0.127
(1.33)
0.073
(0.82)
-0.013
(-0.18)
0.099
(1.11)
0.014
(0.11)
0.068
(0.77)
-0.123
(-0.80)
-0.156
(-1.31)
-0.116
(-1.05)
-0.028
(-0.31)
-0.105
(-0.93)
-0.042
(-0.27)
-0.095
(-0.87)
N
All games
Wins
Losses
Win-Loss
Expected game outcome
Wins
472
-0.176
(-1.41)
-0.197
(-2.18)
-0.178
(-2.33)
-0.090
(-1.16)
-0.170
(-2.00)
-0.264
(-1.95)
-0.179
(-2.13)
Losses
183
0.456
(1.52)
0.498
(2.00)
0.431
(1.77)
0.223
(1.32)
0.556
(2.36)
0.363
(1.28)
0.421
(1.85)
-0.633
(-2.31)
-0.695
(-3.28)
-0.609
(-3.13)
-0.313
(-1.92)
-0.726
(-3.62)
-0.627
(-2.23)
-0.600
(-3.07)
Win-Loss
Unpredictable games
Wins
396
-0.106
(-0.77)
-0.057
(-0.52)
-0.078
(-0.75)
-0.068
(-0.77)
-0.023
(-0.22)
0.102
(0.73)
-0.039
(-0.38)
Losses
338
0.023
(0.14)
0.077
(0.70)
0.024
(0.24)
-0.011
(-0.12)
0.029
(0.28)
-0.003
(-0.02)
0.023
(0.22)
-0.130
(-0.62)
-0.134
(-0.85)
-0.102
(-0.70)
-0.057
(-0.44)
-0.053
(-0.35)
0.106
(0.50)
-0.062
(-0.41)
Win-Loss
34
Table 3. Post-home game day abnormal returns: Highly visible and important games
The table presents the average abnormal stock returns of NFL stadium sponsoring companies in the first
trading day after a home game of their sponsored teams over the period 1997-2013. Abnormal returns are
presented for all home games and separately for wins and for losses of the home teams, within three
samples. The first sample contains all regular season games played on Monday night. The second sample
contains all post-season (playoff) games. The third sample contains all games with unexpected outcome; a
game outcome is classified as unexpected if it is either against the prediction of a pre-game betting spread
of at least 5 points or comes after at least three consecutive games with the opposite outcome. Abnormal
return is the difference between the raw return and the expected return, as measured by the six models
described in Section 3.1. The table also reports the results based on the average of all models. The t-statistic
of the average abnormal return is calculated using the cross-sectional properties of the average. All returns
are in percent and t-statistics are in parentheses.
Interpretation: The outcomes of highly visible and important NFL home games affect the stock price of
the stadium sponsoring companies. Wins in Monday night games have a significant positive effect on
sponsor stock price, an average of 0.51 percent across all models of abnormal return, whereas losses have
virtually no effect (average of 0.01 percent). The difference however between the win and loss returns
shows weak statistical significance (t-statistic of 1.40). Post-season games generate a larger win-loss return
difference of 82 basis points (t-statistic of 1.72), and similarly, games with unexpected outcome yield a
return difference of 81 basis points (t-statistic of 2.50).
N
Meanadjusted
Marketadjusted
Model of abnormal return
Market
Factor Reference Matched Average
model
model
portfolio
firm
Abn return
Monday night games
Wins
52
0.866
(2.07)
0.552
(2.10)
0.211
(0.97)
0.353
(1.82)
0.537
(2.20)
0.530
(1.55)
0.508
(2.18)
Losses
40
0.120
(0.33)
-0.028
(-0.10)
0.071
(0.29)
0.079
(0.32)
-0.081
(-0.31)
-0.093
(-0.21)
0.011
(0.04)
0.746
(1.31)
0.580
(1.52)
0.140
(0.43)
0.274
(0.89)
0.618
(1.71)
0.623
(1.13)
0.497
(1.40)
Win-Loss
Post-season games
Wins
39
-0.353
(-1.48)
-0.198
(-0.80)
-0.284
(-1.27)
-0.246
(-1.01)
-0.241
(-1.11)
0.188
(0.67)
-0.189
(-0.90)
Losses
18
-1.290
(-1.63)
-1.090
(-1.68)
-0.930
(-1.77)
-0.839
(-2.29)
-0.817
(-1.57)
-1.080
(-1.39)
-1.010
(-1.86)
0.937
(1.47)
0.888
(1.56)
0.646
(1.34)
0.593
(1.36)
0.576
(1.21)
1.270
(1.90)
0.819
(1.72)
Win-Loss
Unexpected game outcome
Wins
124
0.821
(2.24)
0.679
(2.05)
0.556
(1.71)
0.222
(1.26)
0.663
(2.07)
0.540
(1.43)
0.580
(1.98)
Losses
154
-0.128
(-0.48)
-0.161
(-0.82)
-0.179
(-1.09)
-0.230
(-1.71)
-0.240
(-1.44)
-0.409
(-1.43)
-0.225
(-1.34)
0.950
(2.13)
0.841
(2.29)
0.734
(2.14)
0.452
(2.07)
0.904
(2.64)
0.949
(2.05)
0.805
(2.50)
Win-Loss
35
Table 4. Pooled time-series regression of abnormal return on home game wins and losses
Using the six models of expected return described in Section 3.1, we estimate the abnormal return for each
of the sponsor-days in the sample. For the full sample of sponsor-days (game and no-game days), we run a
pooled regression of the six-model average abnormal return on two dummy variables: Win indicates a first
trading day after a win of the sponsored home team, and Loss is the equivalent indicator for a loss of the
home team. Panel A shows estimates of OLS regression, and Panel B shows estimates of Generalized Least
Squares (GLS) regression accounting for group-wise clustering by trading day. The results are presented
for the three samples of home games described in Table 3. All coefficients are multiplied by 100 and tstatistics are in parentheses.
Interpretation: The pooled regression estimates corroborate the mean abnormal returns. For all three
samples, the difference between the coefficients of the win and loss dummy variables is very similar to the
win-loss average abnormal return difference reported in Table 3. The GLS regression estimates verify that
the results are not driven by time-clustering.
Monday night
games
Post-season
games
Unexpected
outcome games
Panel A: OLS
Intercept
-0.005
(-1.14)
-0.030
(-5.02)
-0.004
(-0.57)
Win
0.513
(2.37)
-0.159
(-0.54)
0.617
(3.03)
Loss
0.017
(0.08)
-0.977
(-2.26)
-0.223
(-1.22)
Win-Loss
p-value
0.496
(0.049)
0.818
(0.059)
0.840
(0.001)
Panel B: GLS
Intercept
-0.005
(-0.59)
-0.030
(-4.21)
-0.004
(-0.46)
Win
0.539
(2.59)
-0.146
(-0.50)
0.599
(2.95)
Loss
-0.007
(0.05)
-1.027
(-2.38)
-0.236
(-1.30)
Win-Loss
p-value
0.546
(0.015)
0.881
(0.045)
0.835
(0.001)
36
Table 5. Subsequent abnormal returns
Panel A shows the mean difference in the sponsoring companies’ abnormal stock returns in the first trading
day after home game wins and after home game losses (averaged across the six models of abnormal return
described in Section 3.1), and the cumulative average abnormal return during the following four days (days
2 to 5). Panel B shows the slope coefficient estimates of cross-sectional regression of the cumulative
abnormal return (CAR) in days 2 to 5 on the first trading day average abnormal return (AR). The panel also
reports the Wald test p-value for the hypothesis that the coefficient equals -1. Results are presented for the
three samples described in Table 3. Returns are in percent and t-statistics are in parentheses.
Interpretation: Cumulative average abnormal returns in subsequent days do not show return reversal.
Cross-sectional regressions show mixed relations. These results do not support an initial overreaction to
game outcome. Yet the strong relations (positive or negative) between returns in consecutive days may
indicate other forms of behavioral bias.
Panel A: Cumulative average abnormal return
Monday night
Post-season
Unexpected outcome
day 1
days 2-5
day 1
days 2-5
day 1
days 2-5
Wins
t-statistic
0.508
(2.18)
-0.482
(-0.72)
-0.189
(-0.90)
0.182
(0.30)
0.580
(1.98)
-0.252
(-0.58)
Losses
t-statistic
0.011
(0.04)
-0.502
(-0.56)
-1.010
(-1.86)
-2.850
(-1.24)
-0.225
(-1.34)
-0.067
(-0.19)
Win-Loss
t-statistic
0.497
(1.40)
0.021
(0.02)
0.819
(1.72)
3.030
(1.28)
0.805
(2.50)
-0.185
(-0.33)
Panel B: from the regression
2 − 5 = + 1 + Monday night
Post-season
Unexpected outcome
All games
t-statistic
p-value for =-1
0.993
(3.31)
(<0.001)
1.602
(3.53)
(<0.001)
-0.128
(-1.32)
(<0.001)
Wins
t-statistic
p-value for =-1
0.495
(1.24)
(0.001)
-0.489
(-1.05)
(0.278)
-0.251
(-2.03)
(<0.001)
Losses
t-statistic
p-value for =-1
1.674
(3.65)
(<0.001)
2.926
(3.81)
(<0.001)
0.158
(0.93)
(<0.001)
37
Table 6. Post-game trading rule
Every week during the NFL season we form a value-weighted portfolio of buying the stocks of all
sponsoring companies whose teams won that week and selling the stocks of all sponsoring companies
whose teams lost that week. We hold the portfolio from the second trading day after the game until the fifth
trading day. We apply this long-short trading strategy to two samples: all home games and home games
whose outcomes are most unpredictable ex ante (games with betting spreads of 3 points and lower in
absolute value). The table shows the portfolios’ mean excess weekly returns (in excess of the risk-free rate)
and alphas from factor models. The CAPM one-factor model uses the market factor. The three factors in
the 3-factor model are the Fama and French (1993) factors. The 4- and 5-factor models include the FamaFrench factors augmented with momentum and reversal factors. All returns and alphas are in percent per
week and the corresponding t-statistics are in parentheses. The sample period is 1997 to 2013.
Interpretation: Abnormal sponsor stock returns following home games can be translated into profitable
trading opportunities, particularly if applied to the most unpredictable games. The trading strategy generates
mean excess return and factor-model alphas of approximately 28 percent per NFL season.
Excess
return
CAPM
alpha
3-factor
alpha
4-factor
alpha
5-factor
alpha
Sponsors of winning teams
0.577
(2.58)
0.253
(1.52)
0.231
(1.40)
0.234
(1.42)
0.278
(1.70)
Sponsors of losing teams
0.163
(0.66)
-0.191
(-1.03)
-0.204
(-1.11)
-0.187
(-1.01)
-0.139
(-0.76)
Long-short portfolio
0.414
(1.70)
0.444
(1.83)
0.436
(1.79)
0.421
(1.73)
0.417
(1.71)
Sponsors of winning teams
1.525
(3.99)
0.750
(2.51)
0.770
(2.57)
0.782
(2.61)
0.859
(2.88)
Sponsors of losing teams
0.073
(0.17)
-0.750
(-2.13)
-0.746
(-2.11)
-0.730
(-2.07)
-0.603
(-1.72)
Long-short portfolio
1.452
(3.23)
1.500
(3.32)
1.516
(3.34)
1.512
(3.33)
1.461
(3.20)
All home games (N=1,710)
Most unpredictable home games (N=537)
38
Figure 1. Team-specific effect by market size
The sample includes all teams with at least five home wins and five home losses across Monday night
games, post-season games, and games with unexpected outcome (as described in Table 3). The figure shows
the mean difference in the sponsoring companies’ abnormal stock returns after game wins and losses,
averaged across the six models of abnormal return described in Section 3.1. The teams are ordered by the
size of their media market, measured by population in TV households within a 75-mile radius of the team’s
stadium. All returns are in percent.
Interpretation: Most teams show a positive effect of game outcome on the market value of their stadium
sponsoring companies. There is no monotonic relation between the magnitude of the effect and the local
market size of the team.
Difference between sponsor abnormal returns after home wins and losses
2.00
1.50
1.00
0.50
0.00
-0.50
-1.00
Small markets -------------------------------------------------------------------------- Large markets
39
Figure 2. Effect of game outcome on sponsor stock return in home and away games
The upper figure shows the mean difference in the sponsoring companies’ abnormal stock returns after
game wins and losses, averaged across the six models of abnormal return described in Section 3.1. The
lower figure shows the coefficients of the win and loss dummy variables of the regression model described
in Section 3.3. Results are presented for home and away games, applied to the three samples described in
Table 3: Monday night games, post-season games, and games with unexpected outcome. Returns are in
percent and regression coefficients are multiplied by 100.
Interpretation: Outcomes of important home games affect the market value of the team stadiums’
sponsors, whereas away games typically generate a much weaker effect. This suggests that abnormal returns
to the sponsors’ stocks are driven at least to a certain extent by investor sentiment.
Difference between sponsor abnormal returns after wins and losses
1.20
Home games
Away games
1.00
0.80
0.60
0.40
0.20
0.00
Monday night
Post-season
Unexpected outcome
Difference between win and loss regression coefficients
1.20
Home games
Away games
1.00
0.80
0.60
0.40
0.20
0.00
Monday night
Post-season
40
Unexpected outcome
Appendix
Table A1. Publically traded companies sponsoring NFL teams’ stadiums
The table lists all NFL teams’ home stadiums sponsored by publically traded companies over the years 1996-2013.
NFL Team
Bills
Broncos
Browns
Cardinals
Chargers
Dolphins
Eagles
Giants
Jaguars
Jaguars
Jets
Lions
Panthers
Panthers
Patriots
Patriots
Raiders
Raiders
Ravens
Redskins
Saints
Seahawks
Seahawks
Steelers
Texans
Texans
Titans
City
State
Stadium Name
Toronto
Ontario Rogers Centre
Denver
CO Invesco Field at Mile High
Cleveland
OH FirstEnergy Stadium
Glendale
AZ University of Phoenix Stadium
San Diego
CA Qualcomm Stadium
Miami
FL Sun Life Stadium
Philadelphia
PA Lincoln Financial Field
E. Rutherford NJ MetLife Stadium
Jacksonville
FL Alltel Stadium
Jacksonville
FL EverBank Field
E. Rutherford NJ MetLife Stadium
Detroit
MI Ford Field
Charlotte
NC Ericsson Stadium
Charlotte
NC Bank of America Stadium
Foxborough
MA Gillette Stadium
Foxborough
MA Gillette Stadium
Oakland
CA McAfee Coliseum
Oakland
CA Overstock.Com
Baltimore
MD M&T Bank Stadium
Landover
MD FedEx Field
New Orleans
LA Mercedes-Benz Superdome
Seattle
WA Qwest Field
Seattle
WA CenturyLink Field
Pittsburgh
PA Heinz Field
Houston
TX Reliant Stadium
Houston
TX Reliant Stadium
Nashville
TN LP Field
Sponsorship
Years Avg. Annual
Start Date
Sponsoring Firm
Total
Jan 2008
Sep 2001
Feb 2013
Aug 2006
Jan 1997
Jan 2010
Aug 2003
Aug 2011
Jan 1997
Aug 2010
Aug 2011
Aug 2002
Sep 1996
Jan 2004
May 2002
Oct 2005
Jan 2005
Apr 2011
May 2003
May 1999
Oct 2011
Jun 2004
Jun 2011
Aug 2001
Aug 2002
Aug 2010
Jun 2006
Rogers Communications
Invesco Ltd.
First Energy Corporation
Apollo Group
Qualcomm
Sun Life Financial
Lincoln National Corp
MetLife Inc.
Alltel Corp
EverBank Financial Corp
MetLife Inc.
Ford Motor Company
LM Ericsson
Bank of America
The Gillette Company
Proctor & Gamble
McAfee Inc.
Overstock.com
M&T Bank
FedEx
Daimler AG
Qwest
CenturyLink Inc.
H.J. Heinz Company
Reliant Energy
NRG Energy
Louisiana-Pacific
n/a
$ 120,000,000
n/a
$ 154,000,000
$ 18,000,000
$ 37,500,000
$ 139,000,000
$ 400,000,000
$ 6,200,000
$ 16,600,000
$ 400,000,000
$ 40,000,000
$ 20,000,000
$ 140,000,000
n/a
$ 105,000,000
$ 13,700,000
$ 7,200,000
$ 75,000,000
$ 205,000,000
n/a
$ 75,000,000
$ 75,000,000
$ 57,000,000
$ 320,000,000
$ 320,000,000
$ 30,000,000
41
n/a
20
n/a
26
20
5
21
25
10
5
25
20
10
20
15
15
5
6
15
27
10
15
15
20
32
32
10
n/a
$ 6,000,000
n/a
$ 5,923,077
$ 900,000
$ 7,500,000
$ 6,619,048
$16,000,000
$ 620,000
$ 3,320,000
$16,000,000
$ 2,000,000
$ 2,000,000
$ 7,000,000
n/a
$ 7,000,000
$ 2,740,000
$ 1,200,000
$ 5,000,000
$ 7,592,593
n/a
$ 5,000,000
$ 5,000,000
$ 2,850,000
$10,000,000
$10,000,000
$ 3,000,000
Table A2. Game sample distribution
The table presents the yearly distribution of all NFL games (home and away) of teams with publically
traded stadium sponsors. Distributions are presented for the full sample and separately for three subsamples:
regular season games played on Monday night, all post-season (playoff) games, and games with unexpected
outcome, where a game outcome is classified as unexpected if it is either against the prediction of a pregame betting spread of at least 5 points or comes after at least three consecutive games with the opposite
outcome.
Full sample
Monday night
Post-season
Unexpected outcome
# games Home Away # games Home Away # games Home Away # games Home Away
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
31
30
47
49
89
161
195
228
242
275
286
285
262
267
309
319
324
Total 3,399
16
15
23
26
44
80
97
113
128
138
144
144
131
133
157
160
161
15
15
24
23
45
81
98
115
114
137
142
141
131
134
152
159
163
1,710 1,689
1
2
2
5
5
8
8
10
14
18
17
15
16
14
16
18
16
1
2
1
2
2
4
4
5
7
11
8
10
6
5
7
10
7
0
0
1
3
3
4
4
5
7
7
9
5
10
9
9
8
9
1
0
2
4
0
0
2
6
9
14
10
10
15
7
8
11
10
0
0
1
3
0
0
1
5
7
5
5
5
7
3
5
5
5
1
0
1
1
0
0
1
1
2
9
5
5
8
4
3
6
5
2
4
7
18
17
24
27
35
33
55
46
56
44
43
60
43
56
0
1
2
9
12
14
15
14
21
27
22
22
21
24
24
22
28
2
3
5
9
5
10
12
21
12
28
24
34
23
19
36
21
28
185
92
93
109
57
52
570
278
292
42